How to redesign an $800M-a-year software giant for the AI era - Dan Rogers [Asana]

Asana bets the team is now humans plus agents

Asana, eighteen years old with $800m in revenue and 300,000 customers, is betting the team is now humans plus AI agents and that it can be the place those teams work. The reasoning is that models are becoming interchangeable, so durable value sits in context and in a harness that turns model output into governed, repeatable action. AI is already a quarter of net new ARR.

Takeaways

• Asana is building for human-agent teams. Rogers describes the modern team as two people and five agents in a room, and Asana launched Agentic Work Management on the day of recording. Agents offer themselves to users, pick up tasks and write results back, and humans approve the finished work.
• Context and the harness beat the model. Rogers says large language models are becoming interchangeable and the value sits in context, meaning what the company is trying to achieve, who owns the work, past decisions and which systems to use, plus a harness that turns model output into governed, repeatable action.
• AI is a quarter of new ARR. AI teammates and AI Studio went from 17% to 25% of Asana's net new ARR. AI sits in most paid tiers without a price increase, absorbing token costs, and most new deals with Asana's largest customers are AI-attached.
• Agents are billed by the request. Asana created a meter it calls a request, charged when an agent fulfils a requested action rather than by tokens or credits. Every user gets five requests per month to build the habit, which Rogers frames as a freemium sales and marketing investment.
• Stack AI buys speed and a workflow engine. Asana kept the acquired brand as Stack AI by Asana and made it the workflow engine across its products, with hundreds of prebuilt integrations and expertise in pulling information from documents. Acquisitions are meant to shorten delivery, for example from eighteen months to six.
• Asana gains as models improve. Rogers says Asana gets better as models get better and cost per given quality is falling fast. It routes tasks to whichever model performs best rather than the cheapest, and Anthropic is becoming strong distribution while model companies are also Asana customers.

How Rogers took over Asana

1. He arrived with questions rather than answers, aiming to listen twice as much as he talked and to suspend what he already knew.
2. He set out to meet 100 customers in his first 100 days, chosen for diversity of role and account size, including junior people close to the work.
3. He asked how Asana was doing, what problems it solved, how it could do better and what emerging problems customers saw.
4. He sorted existing norms and decisions rather than assuming they were right, discarding some and flagging others for more study.